Few weeks ago i wrote a critical post about Leo, with the new updates i think its fair tto not just bring up negatives but positives too, the overall experience and quality improved a lot its by far the most significant thing i noticed
One person already pointed out bugs i noticed but there is still a few or maybe just “unwanted” features
- On non-claude models skills and instructions get flagged as unsafe regardless of the instruction content, from what i have seen in the glm reasoning it has to do something with the system prompt
- putting images from wikipedia at the start of the convo
Feature requests (in order)
- Some kind of “projects” feature , where you could use the same sources or convo history within that same folder, have some kind of an environment for a topic
- Nicer formatting of responses overall, especially in math it returns either lines of code or just regular text. Im going to be honest i dont know what the specific integration of it but for example chatgpt uses actual math symbols and displays it nicely. With leo i either see dollar signs or just regular (1 2)
- making the output a little wider i think its too narrow
- maybe integrating brave deep research with it somehow?
These are what i had in mind, honestly im very satisfied with leo right now. Look into whatever you guys think is feasible.